LLM Council

关于

The LLM Council is a sophisticated multi-LLM deliberation system designed to tackle complex questions by harnessing the collective intelligence of various large language models. Instead of relying on a single LLM, it employs a multi-stage process: questions are first sent to multiple LLMs in parallel; these LLMs then anonymously review and rank each other's responses to prevent bias; finally, a Chairman LLM synthesizes all inputs into a high-quality, comprehensive answer. This approach ensures more balanced, well-reasoned, and reliable responses by incorporating diverse perspectives and identifying consensus. It is available as a Python library, an MCP server, or an HTTP API.

主要功能

• Multi-stage LLM deliberation system with parallel querying, anonymous peer review, and chairman synthesis
• Secure API key management via system keychain, environment variables, or `.env` files
• Customizable council and chairman LLM models, along with configurable confidence levels for response quality
• Built-in reliability features such as tiered timeouts, partial results, and progress feedback for long-running operations
• 2 GitHub stars
• Support for multiple LLM gateways including OpenRouter, Requesty, and direct provider APIs (Anthropic, OpenAI, Google)

使用案例

• Integrating advanced multi-LLM reasoning capabilities into development environments and other applications via the Model Context Protocol (MCP).
• Obtaining high-quality, synthesized answers to complex or ambiguous questions requiring diverse perspectives.
• Facilitating collaborative decision-making with LLMs, such as approving proposals or reviewing code changes with a binary or tie-breaker verdict.